cancel
Showing results for 
Search instead for 
Did you mean: 

Clustering on AIX 5.3

Former Member
0 Kudos

Dear Friends,

We are facing problem in Failover clustering.

We are going through the implementation phase where in i have setup Production system,

the environment is as follows: SAP (CI) and Database (DB) instance are running on two seperate AIX servers,

We are planning to have cluster in between Ci and DB. That means

1. If Ci fails, it will be mounted on DB

2. If DB fails, it will be mounted on CI.

1. we have tested failover at OS level which is fine.

2. we have also tested Failover of SAP Application manually, which is also working fine.

3. But our main objective s to achive automatic failover between CI and DB, i mean if CI fails it gets mounted and start running on DB automatiacally,

and if DB fails it gets mounted and start running on CI automatiacally.

Please guide me as to how should i go about it...

I kindly request you to give your expert advice....

Regards

Ayush Johri

Accepted Solutions (0)

Answers (3)

Answers (3)

Former Member
0 Kudos

Just want to clear my doubts..

We have a Sun Cluster ( solaris 10) and we tested the failover.. what I have observered is that whenever a service goes down( for eg. ASCS/SCS/DI/CI) it restarts and switch over to next cluster?

So does all the clustering software does the same thing( of restarting )? Then what does the enqueue replication is for? and does it apply to CI /DI? what abt those CI/DI if there is no replication? Then all the services will restart when failover happens?

markus_doehr2
Active Contributor
0 Kudos

We have a Sun Cluster ( solaris 10) and we tested the failover.. what I have observered is that whenever a service goes down( for eg. ASCS/SCS/DI/CI) it restarts and switch over to next cluster?

You mean "next cluster" or "next cluster node"? Yes - it´s the intention of a cluster to keep the application running. I´m not sure what you´re wondering about...

So does all the clustering software does the same thing( of restarting )? Then what does the enqueue replication is for? and does it apply to CI /DI? what abt those CI/DI if there is no replication? Then all the services will restart when failover happens?

Yes - but open transactions will be rolled back and not preserved if you don´t use the enqueue replication.

Markus

Former Member
0 Kudos

hi Ayush,

The automatic failover is the clustering software duty to do it. Whenever a node is not active( or not pingable) the clustering software see this a problem and after the predetermined time( as configured with the cluster s/w) , it will automatic failover to another node.

You need to discuss this with the ibm engineers who has done the clustering thing.... Because it a normal thing in clustering( the automatic failover).

Regards

Former Member
0 Kudos

Hi Ayush,

The scenario you mentioned is standard functioning of cluster. The cluster switchover will happen over the virtual hostname/IP which moves from CI to DB server during failover. You don't need to add anything specific in the startup/stop scripts for this. Your SAP system is refered using the virtual host and will auto direct to DB server when CI failovers to DB.

It is assumed that you have made the relevant changes in SAP profiles for the virtual hostname. SAPLOCALHOST etc.

The scenarios where you will need to alter cluster stop/start scripts is, say in case you run an additional Application instance on the DB server and every time CI fails over onto the DB server you would like to shutdown this Application Instance first befor the CI starts. In this case you will need to add appropriate startup of the Application Instance and shutdown of the Application instance in the cluster start and stop scripts.

There is no need to change the standard sap startsap/stopsap scripts. Hope this clear your doubts.

Thanks and Best Regards,

Sunil.

Former Member
0 Kudos

Hi Sunil,

I appreciate your reply since you precisely answered what i was loking for, but theres soimthng more that i would like to know.

1. Firstly i would like to ask you where can i make changes for the virtual hostname/ip in the sap profile, which profile are you tallking about (start, default or instance).

In ideal scenario we have assinged 10.54.145.32 to CI and 10.54.145.31 to DB

the virtual ips assigned to CI is 10.54.145.40 & to DB is 10.54.145.39

How should i make the changes for these ips and where.

2. please guide me if i am correct or not. while doing manual failover testing (that means we arent using any sap stopsap startsap scripts)

I have mentioned below the procedure of staring SAP and DB when DB fails (please correct if you feel any of the step that i discussed below is wrong)

i fail DB, and DB gets mounted on CI ,(we would now perform all the activities on CI host) so firstly Ci should be stopped with sidadm user (sidadm>stopsap),

then i need to start the listener with orasid user (orasid> lsnrctl start).

Once listner starts i will have to login with sidadm user to start the database (sidadm> startdb) and then again with sidadm i have to start sap application (sidadm> startsap)

what i am trying to ask is during manuall testing, i have to start DB with sidadm user not with orasid (am i right)

Regards

Ayush Johri

Former Member
0 Kudos

Hi Ayush,

I am not sure if I understand you well.

You are trying to do a Failover testing by simulating a failover. Your SAP is configured correctly in the cluster. If this is the case then when you fail the DB, the CI node will automatically recognize the failure of DB node and run the script to acquire/start all resource/application running on the DB.

If your cluster is configured correctly then CI would automatically acquire the virtual hostname/IP, mount the relevant filesystems, start listener service and start the database.

Similarily if you fail the CI node, the DB node would acquire the resources and start Central Instance on the DB node.

This procedure is automated and not manual interference is required is cluster is configured correctly. Trying to do these steps manualy can be quite a pain as you will need to move the virtual IP and hostname, mount the releavnt filesystems ( there maybe some additional steps required here for VGs) and then starting DB/CI manually.

BTW there is no need to stop the CI before starting the DB as the CI would be in reconnect mode till the time DB is up again. (there are obvoisuly few SAP parameter that set the timeout period for this)

The change to SAP profile is in DEFAULT.PFL

rdisp/mshost

SAPLOCALHOST

SAPLOCALHOSTFULL

icm/host_name_full (Instance profile)

are some on the parameters that undergo changes. I would encourage you to go thru the RedBook for setting up SAP in cluster on AIX to understand further. It is quite straight forward.

http://www.redbooks.ibm.com/abstracts/sg244498.html - Chapter 9.

and yes DB is started using sidadm user (as is done by the startsap script)

Thanks and Best Regards,

Sunil.

Former Member
0 Kudos

HI Sunil,

Thanks for this generous explanation, it has been very helpful.

You have understood my problem, but you are trying to explain it from the perspective where we use automated scripts to work with cluster with HACMP, rather i am talking about manual testing.

By manual testing i mean i havent asked my Hardware guy to configure HACMP, my hardware guy has asked me to test it manually, once the manual testing os over he would start configuring automated clustering (i.e. HACMP)

Now my whole idea is to have successful manual failover testing, now i think you must have understood me. firstly i stop all the application i.e. DB and CI. after this i am manually failing DB and DB file system gets mounted on CI host (since clustering is between CI and DB)

Activities on CI host

1. i log in with ora<sid> and start listner 'lsnrctl start'

2. then i log in with <sid>adm to execute 'startdb'

Now at this point when i execute 'startdb' command i get a message saying --> 'Could not find the executable sqlplus'

I am slightly confused as why its saying so and eventually my database doesnt come up.

So this is the way i am trying to test manual failover in whch i am not able to start my Database which is mounted on CI

and why the database is not starting i have mentioned the reson above because of that 'sqlplus' error

Please guide me as to what should i do so that if i execute 'startdb' command it should executable 'sqlplus'

Regards

Ayush

Former Member
0 Kudos

Hi,

Looks like either your enviroment is not set correctly or the /oracle/SID filesystem has not been mounted.

Your PATH variable should have the path to /oracle/SID/102_64/bin and the folder with sqlplus needs to be available. Either of them is missing.

Thanks and Best Regards,

Sunil.

Former Member
0 Kudos

Ayush,

For oracle to start the Oracle file system should be present locally.

What you can do is to have oracle binaries(oracle home/ cd $ORACLE_HOME) on both the node so that whenever the DI instance failover, the binaries are present on the failover node and it will start automatically.

There might be another ways also... depends ...

Former Member
0 Kudos

Hi friends,

the problem is when i execute startdb with user <sid>adm, it says "Could not find the execiutable sqlplus"

Dont know what to do i am badly stuck.

Regards

AYush

markus_doehr2
Active Contributor
0 Kudos

Sqlplus is part of the full Oracle client. It seems that your application server is trying to start the database (instead of the database server).

Markus

Former Member
0 Kudos

Please login through ora<SID> and try to start the listener first.

and then try running the startdb script.

Even if you are stuck.. try starting the DB manually.

conn / as sysdba

startup

and then try to start sap ...

If everything works fine manually, then investigate the startdb script (or the file permission etc) ..

Regards,

Former Member
0 Kudos

Hi prashant,

The method you have suggested to start sap with ora<sid> user, i do that generally as a workaround. but my motive is not to start with ora<sid> user rather i would like to start database from irp<sid> user.

i mean to say that with the user irp<sid> i should be able to start database with 'startdb' command.

Regards

Ayush Johri

Former Member
0 Kudos

Hi,

The reason why you need to login through orairp( irp your SID) is to start Listener and startdb you can do with irpadm.

I am trying to understand your issue.... you can not able to startdb with irpadm?

Have you checked the permission and usergroup.

because normally you startdb through irpadm only. check check the permission... I think that might be the problem

markus_doehr2
Active Contributor
0 Kudos

> 3. But our main objective s to achive automatic failover between CI and DB, i mean if CI fails it gets mounted and start running on DB automatiacally,

> and if DB fails it gets mounted and start running on CI automatiacally.

So what is your problem if you can do the failover manually?

Markus

Former Member
0 Kudos

Hi Markus,

The problem is we want this failover to be automatic, now by automatic i mean it shouldnt be done manually.

So for that we need to have 'stopsap' and 'startsap' scripts running,

and it should be integrated in the hardware scripts.

But i am not able to understand from where would i get these two scripts, and also how should i keep the sequence of commands.

Please advice

Regards

Ayush

markus_doehr2
Active Contributor
0 Kudos

The "startsap" and "stopsap" scripts come with the SAP kernel.

I suggest you take a look at the IBM redbook how to configure HACMP (Chapter 8):

http://www.redbooks.ibm.com/abstracts/sg244498.html?Open

Markus

Former Member
0 Kudos

Hi Markus,

Thanks for the link, i am going through it.

But i am wondering how can i change the standard 'stopsap' and 'startsap' scripts which i found in the kernel directory (/sapmnt/SID/exe) to make it suit to out automatic failover requirement.

Can you please also suggest me a sequence of commands in starting and stopping SAP Ci and DB instance in failover mode...

Regards

Ayush

Former Member
0 Kudos

Just get your IBM unix engg to configure the automatic failover of SAP within the cluster. He/She should be able to incorporate the startup and stopping of SAP within the cluster start and stop scripts.

Generally they have these scripts already available with them and not much of additional work needs to be done unless you have an additional application instance involved or some such extra configuration.

Thanks and Best Regards,

Sunil.

Former Member
0 Kudos

Hi Sunil,

You said it right, these 'startsap' and 'stopsap' scripts are available with the IBM guys, since my engineer was aware of this i asked him to talk to his senior person... who guided him accordingly...

So that issue has been somehow resolved.

But i would like to ask you somthng, regarding the Application servers you said somthng like we need to perform some settings.

So how could these settings be done in 'stopsap' and 'startsap' scripts.

Take an example that my Ci fails, the it gets mounted on DB server,

since my CI has moved to DB server i want Application servers to connect to DB server instead of CI(since CI is down)

In our scenario we have two application servers.

Your advice would be highly appreciated.

Regards

Ayush

SAP Netwevaer